🤖 AI & MACHINE LEARNING

Detect & Correct AI Bias

ZPL gives every AI response a neutrality score from 0.0 to 1.0. Filter biased outputs, balance training data, and monitor model drift — all with one API.

Try ZPL AI Chat Analyze Any Text

Live Demo

AIN Score Bias Meter

Type any text and see the ZPL AIN score in real time. Works in any language.

AIN SCORE

0.0 Biased0.5 Neutral1.0 Perfect

—

Type something to analyze...

AIN Score Reference

0.7 – 1.0 ZPL CERTIFIED

0.4 – 0.7 BORDERLINE

0.0 – 0.4 BIASED

Example Scores

0.95"Climate change has both human and natural causes"

0.55"Electric cars are generally better for the environment"

0.18"This product is absolutely the best, no downsides"

Applications

ZPL Across AI & ML Pipelines

🔍

Response Filtering

Every AI response gets an AIN score. Auto-retry if biased. Works with any LLM via the ZPL AI Proxy.

analyze / rebalance / strict

📦

Training Data Balance

Use ZPL /compute to generate balanced binary labels for imbalanced datasets. Eliminates class bias before training.

Dataset correction

📉

Model Drift Monitoring

Run AIN scoring on model outputs weekly. Track neutrality over time — catch bias drift before it affects users.

Continuous monitoring

📝

Content Moderation

Score user-generated content and AI responses for bias. Flag anything below AIN 0.4 for human review.

Auto-flagging

🌍

Multilingual Bias

AIN scoring works across languages. Romanian, English, Spanish — ZPL AI self-rates neutrality in any language.

Language agnostic

🔗

LLM Comparison

Benchmark different LLMs on neutrality. Which model is most balanced on controversial topics? AIN tells you.

Model benchmarking

Code

Integrate ZPL into Your AI Pipeline

# Python — ZPL bias filter for any LLM output import requests def analyze_bias(text: str, api_key: str) -> dict: """Score any text for bias using ZPL AIN metric.""" res = requests.post( "https://zpl-backend.onrender.com/ai/analyze", headers={"X-Api-Key": api_key}, json={"text": text} ).json() return { "ain_score": res["ain_score"], "certified": res["zpl_certified"], "direction": res["bias_direction"] # positive/negative/neutral } def zpl_filtered_chat(messages: list, api_key: str, mode="rebalance") -> dict: """Chat with Groq AI + automatic ZPL bias correction.""" res = requests.post( "https://zpl-backend.onrender.com/ai/chat", headers={"X-Api-Key": api_key}, json={ "messages": messages, "zpl_options": {"filter_mode": mode} } ).json() return { "response": res["response"], "ain_score": res["zpl_filter"]["ain_score"], "rebalanced": res["zpl_filter"]["was_rebalanced"] }

PRICING

Simple, Transparent Pricing

Run AI bias detection, AIN scoring, and neutrality filtering at scale — from prototype to production.

PRO

$29/mo

Best for developers & researchers

✓ 600 AI analysis calls/month
✓ 35,000 /compute calls
✓ AIN score on every response
✓ POST /ai/chat endpoint
✓ 3 API keys · Max N=25

Get Started